Claude Haiku 5.5 vs Gemini 3.8 Flash: Benchmark Comparison
Claude Haiku 5.5 has the higher score on 17 of 26 shared benchmarks; Gemini 3.8 Flash leads on 9.
The largest observed score gap is 41.00 pts on ProofBench v1.1 , where Claude Haiku 5.5 leads.
Reported ±1 standard-error ranges overlap on 14 of 26 shared benchmarks with comparable uncertainty data. This is not a pairwise statistical significance test.
Shared benchmark results
Scores and reported standard errors come from the latest visible Vals benchmark version available for both models.
| Benchmark | Claude Haiku 5.5 | Gemini 3.8 Flash | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Finance Index | 57.45% ±1.80 | 60.46% ±1.45 | 3.01 pts | Reported ±1 SE ranges overlap |
| Vals Index | 58.71% ±1.28 | 55.95% ±1.14 | 2.76 pts | Reported ±1 SE ranges do not overlap |
| Vals Legal Index | 70.27% ±1.83 | 69.78% ±1.78 | 0.48 pts | Reported ±1 SE ranges overlap |
| Vals Science Index | 38.50% ±1.67 | 35.76% ±1.66 | 2.74 pts | Reported ±1 SE ranges overlap |
| HLAB-Vals-1.1 | 4.17% ±1.52 | 12.08% ±2.58 | 7.92 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 43.27% ±3.44 | 38.94% ±3.39 | 4.33 pts | Reported ±1 SE ranges overlap |
| EMB v1.1 | 65.48% ±3.69 | 72.87% ±2.42 | 7.39 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent v2.1 | 65.37% ±2.06 | 65.76% ±2.04 | 0.39 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 31.92% ±3.35 | 32.12% ±3.37 | 0.20 pts | Reported ±1 SE ranges overlap |
| MedCode | 51.54% ±2.17 | 48.13% ±2.18 | 3.41 pts | Reported ±1 SE ranges overlap |
| MedScribe | 89.54% ±1.92 | 84.50% ±1.94 | 5.05 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 89.00% ±3.15 | 48.00% ±5.02 | 41.00 pts | Reported ±1 SE ranges do not overlap |
| BioMysteryBench | 70.37% ±1.33 | 62.22% ±1.70 | 8.15 pts | Reported ±1 SE ranges do not overlap |
| MysteryMechanism | 35.13% ±3.21 | 36.49% ±3.24 | 1.35 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench-Science | 10.00% ±3.61 | 8.57% ±3.37 | 1.43 pts | Reported ±1 SE ranges overlap |
| SAGE | 45.78% ±3.42 | 35.06% ±3.36 | 10.72 pts | Reported ±1 SE ranges do not overlap |
| Code Migration v1.1 | 52.36% ±3.49 | 38.65% ±3.65 | 13.71 pts | Reported ±1 SE ranges do not overlap |
| IOI | 47.28% ±10.72 | 56.94% ±2.26 | 9.67 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 2.50% ±1.11 | 1.00% ±0.70 | 1.50 pts | Reported ±1 SE ranges overlap |
| SWE-bench Multimodal | 43.96% ±2.27 | 40.83% ±2.25 | 3.13 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench 4.0 | 36.36% ±2.31 | 19.70% ±2.31 | 16.67 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench 1-100 | 20.41% ±3.79 | 18.77% ±3.31 | 1.63 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 90.44% ±1.73 | 78.65% ±3.88 | 11.79 pts | Reported ±1 SE ranges do not overlap |
| CyberBench v1.1 | 75.48% ±5.21 | 43.75% ±2.21 | 31.73 pts | Reported ±1 SE ranges do not overlap |
| SRE Bench | 5.72% ±1.44 | 11.07% ±1.94 | 5.34 pts | Reported ±1 SE ranges do not overlap |
| Public Benefits Bench v1.1 | 63.19% ±1.25 | 65.29% ±1.24 | 2.10 pts | Reported ±1 SE ranges overlap |
Performance by category
| Category | Claude Haiku 5.5 average | Gemini 3.8 Flash average |
|---|---|---|
| Legal | 23.72% | 25.51% |
| Finance | 54.26% | 56.92% |
| Healthcare | 70.54% | 66.32% |
| Math | 89.00% | 48.00% |
| Science | 38.50% | 35.76% |
| Education | 45.78% | 35.06% |
| Coding | 41.90% | 36.36% |
| Cyber | 40.60% | 27.41% |
| Social Mobility | 63.19% | 65.29% |
Cost and latency
Cost per test appears only where the benchmark reports it for both models. Latency is the measured completion time for that benchmark.
| Benchmark | Claude Haiku 5.5 cost | Gemini 3.8 Flash cost | Claude Haiku 5.5 latency | Gemini 3.8 Flash latency |
|---|---|---|---|---|
| Vals Finance Index | $1.39 | $3.70 | 33m31s | 7m44s |
| Vals Index | $3.17 | $6.06 | 49m48s | 35m58s |
| Vals Legal Index | $2.35 | $2.05 | 58m22s | 9m15s |
| Vals Science Index | $1.86 | $2.98 | 39m58s | 22m00s |
| HLAB-Vals-1.1 | $4.26 | $3.66 | 43m41s | 19m22s |
| Legal Research Bench | $1.63 | $1.63 | 1h22m | 5m24s |
| EMB v1.1 | $1.36 | $8.24 | 19m56s | 12m49s |
| Finance Agent v2.1 | $1.64 | $1.99 | 31m05s | 7m24s |
| Tax Agent Bench | $1.16 | $0.86 | 49m34s | 2m58s |
| MedCode | N/A | N/A | 3m46s | 42.89s |
| MedScribe | N/A | N/A | 5m23s | 21.14s |
| ProofBench v1.1 | $0.17 | $0.60 | 6m58s | 6m09s |
| BioMysteryBench | $0.51 | $1.55 | 12m32s | 5m27s |
| MysteryMechanism | $0.54 | $1.73 | 13m25s | 5m22s |
| Terminal-Bench-Science | $4.52 | $5.64 | 1h34m | 55m11s |
| SAGE | N/A | N/A | 2m23s | 28.64s |
| Code Migration v1.1 | $6.20 | $15.79 | 1h26m | 2h35m |
| IOI | $0.83 | $3.98 | 12m34s | 14m20s |
| ProgramBench | $6.83 | $10.93 | 1h13m | 46m39s |
| SWE-bench Multimodal | $0.70 | $3.69 | 16m16s | 22m18s |
| Terminal-Bench 4.0 | $3.12 | $8.65 | 51m56s | 1h47m |
| Vibe Code Bench 1-100 | $8.30 | $18.91 | 4h06m | 1h06m |
| Vibe Code Bench v1.1 | $6.07 | $6.87 | 34m36s | 8m39s |
| CyberBench v1.1 | $1.25 | $1.31 | 32m39s | 9m49s |
| SRE Bench | $3.14 | $27.46 | 57m56s | 2h28m |
| Public Benefits Bench v1.1 | $0.53 | $1.01 | 36m32s | 3m21s |
Results available only for Claude Haiku 5.5
- ClaimBench
Results available only for Gemini 3.8 Flash
- Vals RSI Index
- LegalBench
- MortgageTax
- TaxEval v2
- GPQA Diamond
- MMLU Pro
- MMMU Pro
- LiveCodeBench
- SkillsBench
- SWE-bench
- CUA-bench
- Time Horizon Index: KSP